The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

Summary

Search Update Pipeline: we're resolving issues as they come up in deployment. We still don't have a stable data pipeline, but we're getting there.

WDQS graph split: We've started investigating the different test framework option so that we can validate the split on a subset of queries. The Graph split itself needs a bit more work on some corner cases, but we're getting there. Collaboration with WMDE is working well, we're getting support in terms of analysis.

What we've accomplished

Improve multilingual zero-results rate

work started on repair multi-script tokens split by the ICU tokenizer - https://phabricator.wikimedia.org/T332337

Search Update Pipeline

Number of fixes and improvements to get production ready:
- Filter/drop canary events from update stream - https://phabricator.wikimedia.org/T349591
- logging: add correlating information - https://phabricator.wikimedia.org/T348211
- pick kafka consumer offset defaults - https://phabricator.wikimedia.org/T348112
- Use flink's AsyncRetryStrategy instead of custom retry logic - https://phabricator.wikimedia.org/T347545
- Fetch: Handle Timeout of AsyncAwaitOperator - https://phabricator.wikimedia.org/T347543

WDQS graph splitting

Initial data load of the full graph on test servers is still in progress (started on 2023-10-24). Roughly 70% done so far, but we expect the process to slow down over time - https://phabricator.wikimedia.org/T347504
We will be hosting a session about Graph Split at Wikidata Data Modeling Days -https://www.wikidata.org/wiki/Wikidata:Events/Data_Modelling_Days_2023#Sessions
Initial version of the graph split code, needs more work to support corner cases - https://phabricator.wikimedia.org/T347989

Operations

Java security restarts https://phabricator.wikimedia.org/T350703

Misc

rdf-streaming-updater migration: now running in staging https://phabricator.wikimedia.org/T349095
Ensure mjolnir can work on Python 3.9 or later (as part of the migration to Debian Bullseye) - https://phabricator.wikimedia.org/T346373
Some of our CI pipelines have been migrated to Java 11 - https://phabricator.wikimedia.org/T350587